Computational imaging compressively encodes high-dimensional scene data into low-dimensional measurements and recovers the high-dimensional scene information using computational reconstruction techniques. In the era of big data, the increasing demands for high spatiotemporal resolution have promoted the development of large-scale reconstruction algorithms with high accuracy, low complexity, and flexibility for various imaging systems. The existing large-scale computational reconstruction methods, including alternating projection, deep image prior, and plug-and-play optimization methods, have made great progresses over the past decades. Among the abovementioned methods, the alternating projection has been utilized in gigapixel quantitative phase imaging systems. Besides, the deep image prior and plug-and-play optimization techniques combine the advantages of conventional optimization and deep learning, which hold great potential for large-scale reconstruction. This work reviews the architectures and applications of these methods and prospects for the research trends, which can provide highlights for future works of large-scale computational imaging.